utmpx: fall back to systemd-logind on Linux when traditional utmp is empty/unavailable

[mattsu2020]

On Linux systems with a missing or empty default utmp file, login records can be unavailable to users, who, pinky, and uptime. Try systemd-logind in that case, loading libsystemd at runtime and falling back to traditional utmp if it is unavailable. This replaces the opt-in feature.

Read the boot time from /proc/stat instead of the systemd random-seed timestamp.

Related to #3219.

[mattsu2020]

Is this dynamically linked againsyt libsystemd.so? If so, this should be disabled by default.

Change to fallback only

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/date/date-locale-hour (fails in this run but passes in the 'main' branch)
Note: The gnu test tests/tail/tail-n0f is now being skipped but was previously passing.

[github-actions]

GNU testsuite comparison:

Note: The gnu test tests/pr/bounded-memory is now being skipped but was previously passing.
Note: The gnu test tests/rm/many-dir-entries-vs-OOM is now being skipped but was previously passing.
Congrats! The gnu test tests/tail/pipe-f is now passing!
Skip an intermittent issue tests/cut/bounded-memory (was skipped on 'main', now failing)

[github-actions]

GNU testsuite comparison:

GNU test failed: tests/tail/tail-n0f. tests/tail/tail-n0f is passing on 'main'. Maybe you have to rebase?
Skip an intermittent issue tests/tail/inotify-dir-recreate (fails in this run but passes in the 'main' branch)
Skipping an intermittent issue tests/pr/bounded-memory (passes in this run but fails in the 'main' branch)
Note: The gnu test tests/cut/bounded-memory is now being skipped but was previously passing.
Note: The gnu test tests/rm/many-dir-entries-vs-OOM is now being skipped but was previously passing.
Note: The gnu test tests/seq/seq-epipe is now being skipped but was previously passing.
Note: The gnu test tests/env/env-signal-handler was skipped on 'main' but is now failing.

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/date/resolution (fails in this run but passes in the 'main' branch)

[codspeed]

Merging this PR will degrade performance by 5.46%

⚠️ 126 benchmarks spent significant time in system calls

System calls cannot be consistently instrumented, so they are not included in the measure, which understates the real cost. Please switch to the Walltime instrument to accurately measure system calls.

Measurement and system calls

• cksum_multiple_files
• du_deep_tree[(100, 3)]
• factor_multiple_u64s[2]
• cksum_raw_output
• cksum_crc
• cksum_default
• du_wide_tree[(5000, 500)]
• dd_copy_partial
• wc_bytes_synthetic[1]
• sort_ascii_only[500000]
• …and 116 more

ℹ️ 1 benchmark has no flame graph

The flamegraph could not be generated because too much time was spent in runtime internals, making the visualization not useful for performance analysis.

• sort_dictionary_order[500000]

❌ 2 regressed benchmarks
✅ 399 untouched benchmarks
⏩ 54 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	three_39_bit_primes	583.2 ms	617.8 ms	-5.6%
❌	Simulation	five_38_bit_primes	1.6 s	1.7 s	-5.33%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing mattsu2020:users (aa176f3) with main (6292c03)}

Footnotes

1. 54 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[oech3]

Is there any way to achieve full feature without runtime strong libsystemd.so dependency?
I think libselinux.so is not serious, but this should be part of binaries published at here.
Just parsing systemd files manually?

[oech3]

Or can we believe soname compatibility of systemd and publish them?

[mattsu2020]

Is there any way to achieve full feature without runtime strong libsystemd.so dependency? I think libselinux.so is not serious, but this should be part of binaries published at here. Just parsing systemd files manually?

Since I changed it to dynamic load, I'll test to see if this works.

[oech3]

And remove feat_systemd_logind if succeed.

[oech3]

It seems pure Rust solution is possible: https://github.com/lucab/zbus_systemd (thought maintainer might not want to add such dep...)

[HackingRepo]

nice, ok

[mattsu2020]

Code changes are mostly complete

[oech3]

An additional reason that we want to have this PR: #13212

[Ecordonnier]

It seems pure Rust solution is possible: https://github.com/lucab/zbus_systemd (thought maintainer might not want to add such dep...)

See #8376 (comment) where I explained why I did not use zbus in the implementation of systemd-logind. It is a huge crate.

[Ecordonnier]

Will it still be compatible with GNU coreutils behavior? how much is it deviating? the goal of the initial implementation of systemd-logind was to have the same behavior than GNU coreutils.

[oech3]

@Ecordonnier How about using zbus keeping feat_systemd_logind optional? Then people can keep binary size by removing it.

[mattsu2020]

Will it still be compatible with GNU coreutils behavior? how much is it deviating? the goal of the initial implementation of systemd-logind was to have the same behavior than GNU coreutils.

Mostly, but not completely.

The systemd-derived records are intended to match GNU coreutils, including boot records and separate seat/TTY entries. However, there are still some known differences, such as session PID handling, remote user formatting, and TTY fallback behavior.

[Ecordonnier]

@Ecordonnier How about using zbus keeping feat_systemd_logind optional? Then people can keep binary size by removing it.

Using zbus would grow the binaries by 800KB. I don't think we want this.

[oech3]

I think it is worth to inc multicall-binary 800KiB for simplicity and consistency of binaries published at here.

[oech3]

@Ecordonnier Are you fine to merge this?